This paper describes a physical model of the temporal changes that occur in vegetated land surfaces observed by a repeat-pass radar interferometer. We assume the temporal changes to be caused by a Gaussian-statistic motion of the vegetation elements, with motion variance changing along the vertical direction. We show that the temporal correlation between two interferometric radar signals is affected by the structural parameters of the vegetation, such as canopy height, and varies with the wave polarization. We validate the model using L-band data acquired by the Jet Propulsion Laboratory with the Uninhabited Aerial Vehicle Synthetic Aperture Radar airborne radar. This work provides new insights into the role of temporal decorrelation in interferometric radar applications.